Development of a specific affinity-matured exosite inhibitor to MT1-MMP that efficiently inhibits tumor cell invasion in vitro and metastasis in vivo

Abstract

The membrane-associated matrix metalloproteinase-14, MT1-MMP, has been implicated in pericellular proteolysis with an important role in cellular invasion of collagenous tissues. It is substantially upregulated in various cancers and rheumatoid arthritis, and has been considered as a potential therapeutic target. Here, we report the identification of antibody fragments to MT1-MMP that potently and specifically inhibit its cell surface functions. Lead antibody clones displayed inhibitory activity towards pro-MMP-2 activation, collagen-film degradation and gelatin-film degradation, and were shown to bind to the MT1-MMP catalytic domain outside the active site cleft, inhibiting binding to triple helical collagen. Affinity maturation using CDR3 randomization created a second generation of antibody fragments with dissociation constants down to 0.11 nM, corresponding to an improved affinity of 332-fold with the ability to interfere with cell-surface MT1-MMP functions, displaying IC50 values down to 5 nM. Importantly, the new inhibitors were able to inhibit collagen invasion by tumor-cells in vitro and in vivo primary tumor growth and metastasis of MDA-MB-231 cells in a mouse orthotopic xenograft model. Herein is the first demonstration that an inhibitory antibody targeting sites outside the catalytic cleft of MT1-MMP can effectively abrogate its in vivo activity during tumorigenesis and metastasis.

Keywords: MT1-MMP; antibody; in vivo targeting; metastasis; non-catalytic sites.

Figure 1. MT1-MMP scFv's selected for the ability to prevent collagen cleavage

(A) Aliquots of ectodomain MT1-MMP (50, 20, 10, 5 or 0 nM) were incubated with fibrillated collagen type I for 16 hours at room temperature. The reaction products were analyzed by reducing SDS-PAGE. Collagen cleavage was observed as the conversion of the M_r ∼100 kDa bands of collagen type I to the M_r∼75 kDa ¾-fragments. (B) In separate reactions, ectodomain MT1-MMP was pre-incubated with E.coli. supernatant containing scFv-antibodies selected to MT1-MMP. As indicated by stars, some scFv's were identified as possible hits for interfering with MT1-MMP catalyzed collagen cleavage. In lanes indicated by numbers, MT1-MMP was pre-incubated with control Fc-scFv DES (1), with CT1746 (2) or sample without MT1-MMP (3).

Figure 2. MT1-MMP scFv's selected for the ability to interfere with interactions to TIMP-2

Selected scFv's were screened for the ability to compete with TIMP-2 for MT1-MMP binding. The activity of ectodomain MT1-MMP towards a small fluorogenic peptide substrate was inhibited by TIMP-2, but not by selected scFv's (scFv 1 and 2). However, the TIMP-2 inhibition could be competed away by some scFv's (scFv-1), but not by others (scFv-2), as indicated in the figure.

Figure 3. Lead Fc-scFv's inhibit cell-surface MT1-MMP gelatinolytic and collagenolytic activity

HT-1080 cells were cultured on fluorescently-labeled gelatin (A) or collagen film (B) in the absence or presence of purified MT1-MMP Fc-scFv's, DES control or GM6001. For gelatinolytic activity, cells were fixed after 18 hours and imaged. Staining of actin ensured similar cell numbers in areas of imaging. For collagenolytic activity, cells were fixed after 3 days of incubation and collagen stained with Coomasie Brilliant Blue. As exemplified by Fc-scFv E3, a few clones were able to inhibit both gelatin and collagen degradation. Control Fc-scFv DES served as negative control, whereas GM6001 gave complete inhibition (A and B).

Figure 4. Fc-scFv's inhibit activation of pro-MMP-2, but not activation of pro-MT1-MMP

(A) For pro-MMP-2 activation assays, HT-1080 cells were cultured in the presence of trace amounts of collagen type I in the absence or presence of MT1-MMP Fc-scFv's or GM6001. Following 18 hours of incubation, cell culture supernatant was analysed for levels of pro-MMP-2 and active MMP-2 by zymography. A few lead Fc-scFv's displayed inhibitory activity as shown by E3 and G1, while the majority did not, as exemplified by E10. The experiments shown are representative of a total of three independent experiments. (B) Full ectodomain Pro-MT1-MMP was incubated with or without furin, and in the presence or absence of DES Fc-scFv control or E3, as indicated, for 3 hours at RT. Reaction products were analyzed by reducing SDS-PAGE, and conversion of pro-MT1-MMP to MT1-MMP by furin was observed. (C) Analysis of HT1080 cell lysates for the presence of MT1-MMP was carried out by Western blotting using anti-MT1-MMP polyclonal antibody N175/6. As a loading control, actin was detected by rabbit anti-actin polyclonal antibody.

Figure 5. Fc-scFv's compete with collagen type I for binding to MT1-MMP whereas binding of collagen peptide is unaffected

In an ELISA setup, pepsin-extracted collagen type I or collagen peptide were coated in wells after which ectodomain MT1- MMP was allowed to bind in the absence or presence of Fc-scFv E3 or DES control antibody at a concentration of 0-500 nM.

Figure 6. Affinity maturation of lead antibody clones and screening for improved variants

Homogeneous time-resolved fluorescence (HTRF), TIMP-2 competition and surface plasmon resonance assays were utilized to identify affinity improved variants of lead Fc-scFv's. (A) In HTRF assays, parent E3 binding to MT1-MMP was observed by direct labeling of E3 and detection with labeled streptavidin and biotinylated full ectodomain pro-MT1-MMP. (B) The ability to outcompete labeled parent Fc-scFv E3 from binding to MT1-MMP was detected as a decrease in HTRF signal and displayed as percent of total fluorescence in the absence of competing scFv. (C) Cleavage of small fluorogenic peptide by ectodomain MT1-MMP was inhibited by TIMP-2. E3 Fc-scFv E3 outcompeted TIMP-2 inhibition of MT1-MMP, as shown by increasing E3 concentrations from 0–400 nM. (D) The efficacy of improved variants to outcompete TIMP-2 from ectodomain MT1-MMP was directly compared to E3 parent.

Figure 7. Affinity-improved Fc-scFv's effectively inhibit cell-surface MT1-MMP activity

(A) HT-1080 cells were cultured in the presence of trace amounts of collagen type I, and in the absence or presence of MT1-MMP Fc-scFv's (0–400 nM), DES control (400 nM) or GM6001 (10 μM). Following 18 hours of incubation, cell culture supernatant was analyzed for levels of pro-MMP-2 and active MMP-2 by zymography. Quantification of active MMP-2 levels was done by densitometric analysis of inverted images. (B) HT-1080 cells were cultured on fluorescent-labeled gelatin in the absence or presence of MT1-MMP Fc-scFv's (0–200 nM), DES control Fc-scFv DES (200 nM) or GM6001 (10 μM). Cells were fixed after 18 hours and imaged. Quantification of gelatin degradation was done by densitometric analysis of inverted images. (C) HTC-75 cells were cultured in the absence or presence of MT1-MMP Fc-scFv's (0–1000 nM), DES control (1000 nM) or GM6001 (10000 nM). Following 15 hours of incubation, cell culture supernatant was analyzed for levels of CD44 shedding by ELISA. (D) Effect of Fc-scFv E2-C6 (0–500 nM), Fc-scFv DES control (500 nM) or GM6001 (10 μM) in a collagen invasion assay in which HT-1080 cells were allowed to invade through the collagen matrix for 24 hours towards cell medium containing fetal calf serum as attractant.

Figure 8. Anti-MT1-MMP A4_7 and E2_C6 Fc-scFv's inhibit primary tumor growth and spontaneous metastasis development in mice

MDA-MB-231 LUC2 cells (106) were transplanted into the mammary fat pad of CB-17 SCID mice. A4-7 Fc-ScFv, E2_C6 Fc-ScFv or control DES Fc-ScFv (all at 5 mg/kg) were administered at day four and subsequently three times per week for 10 weeks. (A) Calculated tumor volumes after seven weeks were compared between the three treatment groups using ANOVA and Turkeys multi-comparison tests, (**p < 0.01, ***p < 0.001). (B) Cross sections of the primary tumors from the three different treatment groups. (C) After surgical removal of the primary tumors, the animals were imaged for potential lung and axillary lymph node metastasis development by measuring luciferase activity in the thoracic region using an IVIS Spectrum instrument. (D) Endpoint measurements (10 weeks after initiation of treatment) of photon radiance per area were compared between the three treatment groups using the Kruskal-Wallis and Dunn's multi-comparison test (*p < 0.05). (E) Lung metastasis development was compared between the three treatment groups by calculating the relative metastasis area immunohistochemically stained for human vimentin of randomly chosen sections of mouse lungs. (F) Immunohistochemical stainings of mouse lungs from the three treatment groups showing regions with the most extensive metastasis development. Arrows depict metastatic foci.